BUG: numpy 2.1.0+ raises 'NumPy internal error in AWS lambda #30200
Description
Describe the issue:
I have a simple Lambda function using pyarrow and numpy. The import call results in a call stack to arrayprint.py, which raises this exception:
NumPy internal error: unable to get format_options context variable
If I pin to a version less than 2.1.0, then I don't see this error. I'm guessing this change in 2.1.0 is the source of the regression. I don't know much about the Lambda execution environment, but I've seen issues before with running asyncio loops that required silly workarounds, so perhaps something similar is needed here.
Reproduce the code example:
import pyarrow as paError message:
import pyarrow as pa
File "/var/task/pyarrow/__init__.py", line 61, in <module>
import pyarrow.lib as _lib
File "/var/task/numpy/__init__.py", line 679, in <module>
_core.getlimits._register_known_types()
File "/var/task/numpy/_core/getlimits.py", line 236, in _register_known_types
float128_ma = MachArLike(ld,
File "/var/task/numpy/_core/getlimits.py", line 57, in __init__
self._str_eps = self._float_to_str(self.eps)
File "/var/task/numpy/_core/getlimits.py", line 121, in _float_to_str
return self.params['fmt'] % array(_fr0(value)[0], self.ftype)
File "/var/task/numpy/_core/arrayprint.py", line 1719, in _array_str_implementation
return _guarded_repr_or_str(np.ndarray.__getitem__(a, ()))
File "/var/task/numpy/_core/arrayprint.py", line 571, in wrapper
return f(self, *args, **kwargs)
File "/var/task/numpy/_core/arrayprint.py", line 1701, in _guarded_repr_or_str
return str(v)Python and NumPy Versions:
python 3.12
numpy 2.3.4
Runtime Environment:
AWS Lambda, Python 3.12, x86_64 arch
Context for the issue:
No response